René Widera
René Widera
After long time I found a very nice way which is working also on Cuda. I tested it already together with my alpaka2 playground. I have not tested the solution...
By caller, I mean from wherever the unique ID creation is called from. For shared memory creation it will still be required to provide a unique id but instead of...
With C++20 we can most likely remove all macros because we can get the line and file via native C++.
Thanks for opening the issue, we will have a look into it.
fixed with #2302
Please check that the new implementation has no performance degradation compared to the current dev branch.
@mehmetyusufoglu The CMake file for the benchmark is missing to include catch in case you compile without `-DBUILD_TESTING=ON` and only with `-Dalpaka_BUILD_BENCHMARKS=ON` compiling fails due to missing catch2 headers.
~~The output on the terminal should looks like the old output. ~~ ~~This helps external users if they compare this bable stream against other upstream implementations.~~ ~~The user should not...
> > The output on the terminal should looks like the old output. This helps external users if they compare this bable stream against other upstream implementations. The user should...
@mehmetyusufoglu please amend and force push this PR again, the CI is fixed